Icalia Docker Praxis

This is a collection of examples and notes about our Docker practices at Icalia Labs

About Mentors

We recommend you to identify the available platform mentors in the team so whenever you find a problem regarding Docker and the development technology you're using, you know the right person to help you with your issue:

Expertise	Guru	Contact Email
All Things Docker	Vov	[email protected]
Ruby + Rails	Vov	[email protected]
Elixir	???	[email protected]
Python	???	[email protected]
Java	???	[email protected]
PHP	???	[email protected]
ReactJS + Docker	???	[email protected]
EmberJS + Docker	Vov	[email protected]

About our team's workstation OS diversity

We do want our team members to use their operative system of choice (Linux, MacOS or Windows*), whenever it is possible. Each team member is responsible to notify the appropiate mentor about any problem found. The following considerations must be followed to allow collaboration from the rest of our team members, regardless of their workstation OS of choice:

Docker for Mac

General Considerations

• ENSURE THE **/*.DS_Store LINE IS INCLUDED IN THE .gitignore FILE! (and any other ignore files)

Docker for Linux

The File Permissions Issue

We run our development container processes as root in our projects (See Why we run development container processes as root). This practice poses no problems when using Docker for Mac/Windows, as the shared files in the host are replicated to the guest Docker VM without file owner/group data. But that's not the case for our teammates using Docker for Linux, so when generating files inside a container (i.e. when using the rails generate command), the generated files belonged to the host's root user, preventing them from editing those files with their IDE's, and making a poor development experience with Docker.

The best strategy we've found to work around this is to "re-map" the Docker container's root user & group IDs to the host's developer user & group IDs, using a Docker daemon isolation feature (See ability to remap subordinate user and group ids for more info)

These are the steps to configure your Docker daemon this way:

1. Prune all your current Docker images, containers and volumes

Once the remap configuration is active, you'll no longer be able to access (via docker) your existing images, containers and volumes, adding up to your workstation's "unreclaimable" disk space. It's better to prune all of that out:

docker system prune --all --force --volumes

2. Make sure you know your user ID and group ID on your host:

# be sure to put your username here, or omit it:
id testuser
uid=1000(testuser) gid=1000(testuser) ...

3. Next, you'll need to edit the /etc/subuid and /etc/subgid files, which define the id mapping as an offset between the host uid/gid and the container processes' uid/gid. Notice how some linux distros may have these files already created by default. Also, keep in mind that

# at /etc/subuid: match the second column (separated by colons) to your user id
testuser:1000:65536

# at /etc/subgid: match the second column (separated by colons) to your group id
testuser:1000:65536

These entries mean that for the testuser, the subprocess uid will be an offset of the host's uid, and only applies for the following 65536 uids:

• The container's root user, having the 0 uid will be remapped to the host's 1000 uid, which belongs the host's testuser user.
• If there's some other user in the container, having a 500 uid, it will be remapped to the host's 1500 uid, while the container user having the 501 uid will be remapped to the host's 1501 uid.

4. Now, you'll need to configure your host's Docker engine to apply the remapping in the config file at /etc/docker/daemon.json (not present by default):

{
  "userns-remap" : "testuser"
}

5. Be sure to restart your Docker daemon:

sudo service docker restart

6. Finally, test that the files generated within the container have the correct owner on the host ( your user):

# Run it from within your desktop, bind-mounting it into the container:
cd ~/Desktop && docker run --rm -ti -v $(pwd):/my-desktop -w /my-desktop alpine ash

# Once inside the container - be sure no 'permission denied' error appears:
echo "TEST" > test.txt

# Exit the container:
exit

# Once your'e back to your host, check that the created file belongs to your user & group:
ls -lah

If something here didn't go as planned, please check:

• Your user id and group id matches those on the /etc/subuid and /etc/subgid.
• Your'e pointing the "userns-remap" to the corresponding entry at the /etc/docker/daemon.json.
• You have restarted the Docker daemon after configuring the user namespace remapping.

If your'e still having problems, ask for help to corresponding mentor (@vovimayhem).

Docker for Windows

Important Stuff

These are the common problems we've encountered when using Docker for Windows. All of these are not with Docker for Windows itslef, but rather some gotchas found when working with diverse environments (Teams working with Windows and Linux):

• Make sure you are not using "automatic line-ending conversions" See this one

• Avoid having extension-less executable scripts on source code: A common rule fo .gitattributes files to fix the problem is *.sh text eol=lf. Even having this rule, you might still get in trouble if there are scripts without the .sh extension. Ensure all scripts intended to run on linux containers end with *.sh. If you require them to not have the extension inside the container (i.e. restoredb and dumpdb), you can change that with a custom volume mapping:

version: "2.4"

volumes:
  postgres_data:

services:
  postgres:
    image: postgres:10-alpine
    volumes:
    - postgres_data:/var/lib/postgresql/data
    - ./bin/dumpdb.sh:/usr/local/bin/dumpdb:ro
    - ./bin/restoredb.sh:/usr/local/bin/restoredb:ro
    - ./db/dumps:/db/dumps

General Considerations

• Docker for Windows will only work for Windows 10 Pro, which contains the Hyper-V hypervisor.